home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.c++
- Path: sparky!uunet!decwrl!parc!boehm
- From: boehm@parc.xerox.com (Hans Boehm)
- Subject: Re: Garbage Collection for C++
- Message-ID: <boehm.715281775@siria>
- Sender: news@parc.xerox.com
- Organization: Xerox PARC
- References: <DAVEG.92Aug17224359@synaptx.synaptics.com> <1992Aug20.213720.16599@microsoft.com> <boehm.714410546@siria> <1992Aug25.185458.10190@microsoft.com> <boehm.714847027@siria> <TMB.92Aug29175419@arolla.idiap.ch>
- Date: 31 Aug 92 17:22:55 GMT
- Lines: 23
-
- tmb@arolla.idiap.ch (Thomas M. Breuel) writes:
- >To inject some real-world data again, fragmentation can be a problem.
- >I have a machine with about 200M swap space, and a program that only
- >needed about half that much memory ran out of memory because of
- >fragmentation.
-
- Could you give more details as to which allocator this was? Even
- with nonmoving allocators, there is a lot of variation as to how
- much fragmentation they introduce.
- As I've said before, we have yet to see this sort of ratio
- with an allocator that coalesces large objects, and coalesces small objects
- if an entire page is free. And I have seen significant fragmentation
- primarily in situations (multiple very large object allocations) in
- which moving collection would have been expensive.
- Of course, your application may have been different from mine ...
-
- But it certainly is the case that if you can move objects, you have some
- hope of obtaining a provable upper bound on space consumption, something
- you're unlikely to be able to do with either malloc/free or a nonmoving
- collector.
-
- Hans-J. Boehm
- (boehm@parc.xerox.com)
-